Factors Affecting Permanent Partial Disability Ratings in Workers Compensation

Transcription

1 The Journal of Risk and Insurance, 1998, Vol. 65, No. 1, Factors Affecting Permanent Partial Disability Ratings in Workers Compensation David L. Durbin Jennifer Kish ABSTRACT Using a new and rich data source on permanent disability claims, this article explores two issues relating to permanent partial disability (PPD) benefits. First, the process of determining the final disability rating, which affects the actual level of PPD benefits, is modeled. Differences between the initial impairment rating provided by the treating physician and the final disability rating are investigated to determine what factors, in addition to physical impairment, influence the determination of the final permanent, partial disability rating and, hence, the amount of indemnity benefits. This investigation highlights and measures some of the equity concerns inherent in establishing PPD benefits. Second, the influence of the final disability rating on the dollar amount of the permanent partial and total indemnity benefit award is measured. The empirical findings suggest that the same injuries, as measured by a physician impairment rating, may receive very different final disability ratings and thus have very different claim costs as a result of factors unrelated to the injury. These differences are magnified across states. INTRODUCTION The adequacy and equity of benefits for permanently disabled workers continues to be a vexing problem for workers' compensation insurance. The 1972 National Commission on State Workers' Compensation Laws offered the view that permanent partial benefits "are the most controversial and complex aspect of workmen's [sic] compensation cases. The National Commission also observed that for no other type of benefit are there more variations among states or more divergence between statutes and practice. Even after 20 years of significant reforms, these issues remain unresolved today. David L. Durbin is Senior Vice of Economic Research at Swiss Re and Jennifer Kish is an actuary at St. Paul Re. This research was conducted while the first author was Vice President and Director of Research at the National Council on Compensation Insurance. The authors thank David Appel, Richard Butler, Jack Worrall, and participants at the 1995 American Risk and Insurance Association Annual Meeting for helpful comments. Errors remain ours.

2 82 The Journal of Risk and Insurance Moreover, recent studies have suggested that a substantial portion of the significant cost increases in the 1980s into the early 1990s in workers' compensation may be due to higher costs associated with permanent disability cases. Permanent partial disability (PPD) benefits account for approximately twothirds of the total costs of workers' compensation disability benefits (Berkowitz and Burton, 1987). Not surprisingly, the proportion of total benefits accounted for by permanent partial compensation also has been shown to be positively correlated to employer premiums (Krueger and Burton, 1990). Both the frequency and average cost of PPD cases have increased significantly in recent years. During the six-year period from 1988 to 1994, average PPD costs increased 25 percent, an increase of approximately 4 percent per year, while the frequency increased almost 29 percent or 4.4 percent annually. These increasing average costs and frequencies have resulted in an increase in the permanent partial cost per worker of approximately 52 percent (over 7 percent annually) over the same six year period. (National Council on Compensation Insurance, ). Despite the importance of permanent partial benefits as a contributing factor to the increase in workers' compensation costs, little research has examined the factors that influence the costs of permanent partial disability claims. Using a new and rich data source on permanent disability claims, this article explores two issues relating to permanent partial disability benefits. First, we model the process of determining the final disability rating, which affects the actual level of PPD benefits. In this analysis, differences between the initial impairment rating provided by the treating physician and the final disability rating are investigated to determine what factors, in addition to physical impairment, influence the determination of the final permanent partial disability rating and, hence, the amount of indemnity benefits. This investigation highlights and measures some of the equity concerns inherent in establishing PPD benefits. Second, we measure the influence of the final disability rating on the dollar amount of the permanent partial and total indemnity benefit award. The empirical findings suggest that the same injuries, as measured by a physician impairment rating, may receive very different final disability ratings and thus have very different claim costs as a result of factors unrelated to the injury. These differences are magnified across states. After a review of relevant aspects of workers' compensation systems, we discuss theoretical concerns and briefly review the literature. The next two sections discuss data and empirical results. Finally, we conclude with policy implications. CHARACTERISTICS OF WORKERS COMPENSATION SYSTEMS Workers' compensation insurance provides benefits to workers for on-the-job injuries and work-related diseases. These benefits include unlimited medical benefits, wage replacement (or indemnity) benefits, and vocational rehabilitation.

3 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 83 Indemnity benefits are typically classified, according to the nature and degree of disability, into five categories: fatal, permanent total, permanent partial, temporary total, and temporary partial. In general, any worker with an injury resulting in time loss from work (not including medical-only cases) receives either temporary total, or temporary partial disability benefits (benefits paid during the healing process). Temporary benefits are typically paid until the claimant reaches the point of so-called "maximum medical improvement" (MMI) or is given a medical release to return to work (actual practice of this will vary by state). If the worker suffers some permanent impairment as a result of the injury, permanent partial benefits (or permanent total benefits if 100 percent disabled) are typically awarded after MMI has been reached. Most states determine permanent partial benefits based upon whether an injury is "scheduled" or "nonscheduled" (Virginia and Kentucky are exceptions). Scheduled injuries, typically involving limbs, eyes, and ears--are often compensated by an amount based solely on medical impairment, without regard for loss of earning capacity. To arrive at the award, the number of weeks listed in the schedule for each body part is multiplied by the percentage of functional impairment.. In Tennessee, for example, a knee sprain resulting in a 20 percent impairment rating to the leg would receive 40 weeks of benefits (20 percent of the scheduled 200 weeks for loss of a leg) at 66 2/3 percent of the claimant's preinjury wage (subject to state-specified minimum and maximum restrictions). In contrast, nonscheduled injuries are usually injuries to body parts, typically internal organs, the trunk, neck, or head that are not specified in the schedule, as well as psychological claims (classifications vary by state) and are compensated either by an amount based on functional impairment or on predicted loss of earning capacity. For example, a Tennessee claimant given a 20 percent impairment rating to the whole body (either due to physical impairment or loss of earnings) for a back injury, would receive 80 weeks of benefits (20 percent times 400 weeks for disability of the whole person) at 66 2/3 percent of the claimant's preinjury wage (again subject to state restrictions). Where applicable, loss of earning capacity benefits are paid to those claimants who either do not return to work, or return to work at a wage that is less than their preinjury wage. In some states, claimants may choose to have permanent partial benefits awarded based on either physical impairment or projected future (loss of) earnings (Georgia, for example). Other states base awards on physical impairment in addition to wage loss (Arkansas) while some states base awards only on physical impairment (Virginia). The determination of physical impairment is, at least conceptually, fairly straightforward. Physical impairment ratings are awarded by the treating physician, often in accordance with the American Medical Association's Guides to the Evaluation of Permanent Impairment. In contrast, loss of earning capacity rating procedures tends to be much more loosely defined. Many states have no written rules about how to arrive at a rating. Earning capacity ratings are typically determined by comparing the claimant's preinjury earnings with

4 84 The Journal of Risk and Insurance projected future

5 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 85 earnings. A variety of factors may be considered including age, education, experience, previous occupation and earnings, and the claimant's willingness and ability to train for an occupational change. Typically, the awarded degree of permanent disability, either based on impairment or loss of earnings or both (the final disability rating), along with the claimant's preinjury wage, determines the dollar value of the permanent partial or permanent total disability award. However, the methods used to determine permanent disability ratings differ by state, and often include other factors beyond the treating physician's determination of impairment, such as age, educational level, or the predicted future loss of earnings. Quite frequently, the final disability rating and award are codified in a settlement agreement between the injured worker and the employer/insurer. Additional amounts may be added to the final award, generally in non-scheduled cases, as "incentives to settle" (sometimes also called clinchers) or for coverage of any future medical or vocational rehabilitation costs. THEORETICAL CONSIDERATIONS A neutral third party, either the State Workers Compensation Board or the courts theoretically determine the degree of permanent partial disability awarded. In practice, however, the award is often determined by a settlement agreement between the insurer and the claimant. Although in theory, this determination is based on the presence or absence of medical symptoms (i.e., the physician's determination of physical impairment), the claimant may influence the size of a permanent partial disability rating (and hence the permanent partial award) in several ways. These include: (1) claimants must first decide whether to apply for permanent partial benefits, and (2) claimants may seek to influence the compensation board or insurer by investing in resources designed to influence either the board's decision or the insurer's offer. Legal counsel as well as additional medical and vocational rehabilitation evidence may affect the workers' compensation board decision or the willingness of the insurer to settle the claim. The extent of permanent partial disability also may be influenced systematically by a number of factors such as age, sex, pre-injury wage, and preexisting conditions. Stated differently, although in most states statutory language governs the determination of permanent partial disability awards, in practice these awards may be affected by other considerations. For example, the state may either implicitly or explicitly promote income distribution to "poorer" workers through the workers' compensation system. Workers' compensation systems also have appeared to evolve (both statutorily and through case law) in response to moral hazard problems inherent in the system (see Durbin and Butler, 1998) and the rent seeking activities of some of the participants. It is important to note that these theories may not be mutually exclusive: states may seek to redistribute

6 86 The Journal of Risk and Insurance income, and there may be rent seeking. 1 Previous Research Several studies examine the probability that a workplace injury becomes a permanent partial disability claim. Thomason uses a sample of New York compensation claims occurring from 1972 to 1977 to show that the probability of a permanent partial disability claim is positively related to increasing permanent partial benefits. This result, though, is statistically significant for only nonscheduled claims, which in New York, received awards based on wage loss. Worrall et al. (1993) uses data on individual claims from Massachusetts and Illinois to estimate the probability that a claimant who receives temporary total benefits will go on to receive permanent partial benefits. This study indicates that the claimant's preinjury wage is negatively related to the probability of receiving permanent partial benefits, and the level of compensation benefits is positively related, although these results were not always statistically significant. The focus of these studies is measuring the moral hazard implications of rising workers' compensation benefits. There is growing research on the determinants of attorney involvement and litigation in workers' compensation. (Borba and Appel, 1987, and Boden and Victor, 1994, provide representative examples) hypothesizes that workers will retain legal counsel if the expected benefits of doing so outweigh the costs (both monetary and psychic). There is evidence that attorney involvement significantly increases workers' compensation claim costs, although the effect may vary by type of injury (Durbin, Helvacian, and Butler 1995). Litigation or its threat appears to increase total claims costs (including the benefit award) for more objective injuries like fractures, amputations, etc. For more subjective injuries such as sprains and strains, litigation increases the total costs, but much of the increase is for attorney fees; the injured worker may not actually receive any different net award. For more objective injuries, attorneys may serve a negotiation role in the sense of trying to gain larger settlements. For the subjective injuries, attorneys may act more as workers' compensation system navigators assisting workers in receiving awards. Research on the Canadian workers' compensation system (Sinclair and Burton, 1994) takes a slightly different approach. Workers, both healthy and injured, evaluate their quality of life postinjury. These evaluations are then correlated by type of injury with the American Medical Association (AMA) impairment ratings. The results suggest that workers place a higher value (loss) on an injury than is compensated by the impairment ratings. The researchers suggest that noneconomic loss from an injury is significant and undercompensated. However, little previous research has explored factors involved in the determination of the final disability award. Thomason, using a sample of New 1 We are indebted to an anonymous referee for clarifying the alternative theories.

7 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 87 York closed claims occurring between 1978 and 1984, found that, although higher permanent partial benefits increase the probability of a permanent partial disability award, they are negatively related to the size of the award. From these results, Thomason concludes that the claimant can increase the probability of a permanent partial disability award but not the award size. This study also suggests that the process of determining whether a claimant is permanently disabled is fundamentally different from that determining the size of the award. There has been no previous attempt to explore the income redistribution versus rent seeking theories and their potential impact of workers' compensation awards. The Closed Claim Sample RESEARCH METHODOLOGY The data used in this study was collected through the National Council on Compensation Insurance s (NCCI s) Closed Claim Study Initiative. 2 This study examines random samples of high cost claims from major workers' compensation insurers in ten states. 3 The intent is to identify and measure the forces affecting the significant growth in workers' compensation claims costs. The data were assembled through a detailed review of individual claim files; more than 200 pieces of information were collected on each claim. Besides basic demographic information, the data consist of detailed information on the amount and type of benefits, whether litigated or settled, and information on the adjudication process. Workers' compensation insurers were asked to participate in each state based on their share of the market. In addition, the data contain information on medical impairment ratings, whether these ratings are disputed, alternative medical ratings, final disability ratings, and whether final disability ratings are determined based on medical impairment or loss of earnings. In order to facilitate comparisons between states, medical impairment and final disability ratings are converted to a "whole body" basis using the AMA s guidelines. For example, 50 percent impairment rating to the hand translates into 27 percent impairment to the whole body, using the AMA's Guides to the Evaluation of Permanent Impairment. The claims used in the analyses are major indemnity (lost time) cases closing in specifically selected time periods (for each state, two periods are selected -- either 1989 or 1991, or 1990 and 1992). The indemnity thresholds used to select claims vary by state and closure year. Thresholds were chosen to ensure that the percentage of claims meeting the threshold would be roughly constant by state and 2 The NCCI, headquartered in Boca Raton, Florida, provides data base products, software, publications, and consultation services to state funds, independent bureaus, agents, regulatory authorities, and more than 700 insurers. 3 The ten states are Arkansas, Georgia, Kentucky, Missouri, New Hampshire, North Carolina, South Carolina, Tennessee, Virginia, and Wisconsin.

8 88 The Journal of Risk and Insurance closure year. 4 After data editing, the final number of claims for each state ranged from 598 cases for Wisconsin to 796 cases for New Hampshire. A total of 6,297 claims were available for analysis. After restricting the sample to those claims with both a medical impairment rating and a final disability rating, the final database used in the analysis contained 4,265 claims. Descriptive Statistics Table 1 contains the mean values of the medical impairment rating, the final disability rating, and the mean value of the difference between the ratings for each state studied. Despite similar samples of high cost claims, these mean values differ significantly by state. The average final disability rating ranges from 16.6 percent of the whole body in Georgia and Virginia to 39 percent in Tennessee. Similar differences are observed by state for medical impairment ratings. Even more notable is the significant variability between the medical impairment and final disability ratings. Georgia and Virginia exhibit virtually no difference between the two (medical impairment is actually higher), while the final disability rating is more than double the impairment rating in Tennessee. 4 The state thresholds were chosen so that the indemnity portion of the population of claims above the threshold would account for more than 80 percent of the state's total indemnity costs (based on a review of claims reported under the NCCI Unit Statistical Plan).

9 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 89 Table 1 Mean Impairment Ratings, Final Disability Ratings, and Difference Between Ratings by State State Physician Impairmen t Rating Final Disability Rating Difference Between Ratings Percent of Claims with Impairment Rating Percent of Claims with Disability Rating Correlation Between Impairment and Disability Ratings* Arkansas Georgia Kentucky Missouri New Hampshire North Carolina South Carolina Tennessee Virginia Wisconsin Average * All correlation coefficients significant at p <.001 The percentage of sample claims that receives a medical impairment rating from the treating physician also varies by state. Only 34 percent of claims in New Hampshire have an impairment rating given, compared to 91 percent in Arkansas. The percentage of the sample claims with each type of rating is shown in columns 4 and 5 of Table 1. The correlation coefficient between the impairment rating and the final disability rating was determined to be on average across the entire sample (based on claims with both an impairment rating and a final disability rating). This suggests that, while significantly positively correlated, other factors in addition to physical impairment are involved in the determination of final disability. States vary significantly in the correlation between their impairment and disability ratings, further illustrating problems in designing permanent partial disability systems. The lowest correlation (0.346) is observed in Kentucky; the highest correlation is observed in Virginia (0.929). EMPIRICAL ANALYSIS This section uses multivariate regression models to identify and measure factors related to the size of the final permanent partial disability rating and determines

10 90 The Journal of Risk and Insurance how the size of this rating influences the total indemnity cost of the claim. The research plan controls for the nature, severity, and type of injury and observes how other factors may systematically influence the final disability rating and claim costs. Model 1: Size of Final Disability Rating The final disability rating for permanent partial injuries is determined at a point in time, typically either at the date of maximum medical improvement or when a settlement agreement is negotiated. Given the panel nature of the data, the process of determining the final disability rating may be modeled as follows: Y ijt=βx ijt + u j + Γt + e jt (1) where Y ijt = size of the final disability rating (percent of whole body times 100) for person i, in state j, accident year t, β = vector of model parameters, X = vector of variables influencing the rating size, U j = omitted state effects, Γt = omitted effects that vary across time but not across states, and e jt = independent and identically distributed random variable. This fixed effects specification allows the model to be estimated using ordinary least squares and will provide unbiased parameter estimates. However, as Thomason demonstrates in the context of litigation and claim settlement decisions, and since the data are by definition limited to claims with both final disability ratings and medical impairment ratings, the regression models are subject to selectivity bias. In the empirical analysis, Heckman's two-step estimation procedure is used to correct for sample selection bias (see Heckman, 1979 and Greene, 1981 for relevant discussions). This model estimates the effect of selected variables on the size of the final disability rating with the aim of measuring and identifying those factors with the largest impact on the final rating. The dependent variable is the size of the final disability rating. The regressions include the claimant's age, sex, preinjury weekly wage, weekly temporary total indemnity benefit, claimant attorney representation, whether there is a dispute over indemnity or medical/rehabilitation benefits, and whether the final disability rating is based on loss of earnings. 5 The model also includes control variables for the nature of injury, the body part injured, whether surgery is involved, and the extent of medical costs (severity control variables), as well as dummy variables for the state and accident year (the 5 The loss of earnings base variable is a claim-specific dummy variable indicating whether the PPD award was primarily based on the claimant's loss of earning capacity as opposed to physical impairment. Most states allow awards to be based on either loss of earnings or physical impairment.

11 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 91 fixed effects) in which the claim was initiated. The sample means and standard deviations for the selected variables are shown in Table 2. Similar to other workers' compensation databases, most claimants are male, in their mid-to-late 30s, with a large proportion of back injuries (32 percent). Since these claims are mostly permanent disabilities, there are some other notable features. Most of these claims (78 percent) involve attorney representation, many involve surgery (68 percent), most of the claims are closed by agreement (90 percent lump sum), and 31 percent involve a pre-existing condition. The total average cost (indemnity plus medical) of these claims is $58,583. Again, these claims are serious high cost claims, which although atypical, account for a disproportionate share of costs. 6 Consistent with previous research, the models are estimated in natural logarithmic form. In addition, since de jure in many instances the impairment rating should equal the disability rating, or at least be jointly determined, simultaneous equation models are estimated. Two-stage least squares models are used to jointly estimate the impairment and disability rating awards. Statistical tests indicate the appropriateness of considering a jointly determined model. 7 Because similar injuries to older workers are likely to have more serious consequences, we expect age to be positively correlated to the final disability rating. Claimant attorney involvement and disputes over benefits also are expected to be positively correlated to the final disability rating, consistent with the hypothesis that claimants are able to affect their ratings through investing in legal resources and disputing benefits. Claim severity measures should likewise be positively correlated with the disability award. 6 According to the NCCI Unit Statistical Plan Data, permanent partial injuries account for approximately 8 percent of claims and 75 percent of all indemnity costs. 7 The Wu-Hausman test rejects the null hypothesis of separate equations at the level of significance.

12 92 The Journal of Risk and Insurance Table 2 Sample Means and Standard Deviations (N = 4265) Sample Mean Standard Deviation Variable Wage Weekly Benefit Age Male Attorney Representation Dispute Over Indemnity Benefits Dispute Over Medical/Rehabilitation Benefits Loss of Earnings Basis Lump Sum Award Physician's Impairment Rating Final Disability Rating Pre-existing Condition Surgery Head Injury Neck Injury Low Back Injury Trunk Injury Lower Extremity Injury Upper Extremity Injury Multiple Body Part Sprain or Strain Injury Fracture Injury Total Indemnity 39,114 24,101 Total Medical 19,469 19,435 Table 3 contains the estimated coefficients from the simultaneous equation model where the impairment rating and the final disability rating are jointly determined. The estimated parameters are consistent with the predictions, and, overall, the regression model explains roughly 59 percent of the variation in the disability ratings. 8 As expected, older workers received higher disability ratings and impairment ratings: For every 10 percent increase in age, the impairment rating is 2.3 percent higher and the final disability rating is approximately 0.9 percent higher. Men receive 1.7- percent higher impairment ratings and 0.4 percent higher disability ratings. Attorney involvement is associated with higher disability awards, but with lower impairment ratings. Disputed claims also give 8 The regression coefficients may be interpreted as elasticities. For continuous variables that enter the model in the natural logarithmic form, the coefficients represent the percentage change in the outcome measure given a percentage change in the individual explanatory variable. However, for the qualitative or dummy variables that enter the model as either zero (no) or one (yes), the elasticity is measured as e b - 1. See Halvorsen and Palmquist (1984) for discussion on the interpretation of dummy variables in semilogarithmic specifications.

13 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 93 rise to higher disability awards. These results support the notion that many factors, in addition to physical impairment, affect the size of the final disability rating. The estimated coefficient for the impairment rating is 0.748, implying that a 10 percent increase in the impairment rating gives rise to a 7.5 percent increase in the final disability award. This finding, which comes from models that control for the other factors typically considered in settlement proceedings, is statistically different from one. Table 3 Physician s Impairment Rating and Final Disability Rating Two Stage Least Squares Regressions (with Sample Selection Correction) (N = 4265)

14 94 The Journal of Risk and Insurance Variables Physicians Impairment Rating Coefficient t Statistic Final Disability Rating Coefficient t Statistic Physician's Impairment Rating 0.748* Wage ** Weekly Benefit * ** Age 0.232* * Male 0.173* ** * Attorney Representation ** ** Dispute over Indemnity Benefits 0.064** Dispute over Medical/Rehab Benefits 0.070** * Loss of Earnings Base 0.576* Lump Sum Award 0.053** * Total Medical Cost * Surgery Pre-existing Condition * Release to Work Sprain or Strain * Part of Body Controls Yes Yes No No State Controls Yes Yes Yes Yes Accident Year Controls Yes Yes Yes Yes Lambda ** * F statistic 26.6* 207.1* Adjust R * Significant at.01 ** Significant at.05 *** Significant at.10 Thus, as might be expected from reading the statutory intent, there is not a one-toone relationship between impairment and final disability ratings. Moreover, while statistically significant, the adjusted R 2 of.59 indicates the existence of considerable noise or unaccounted-for factors in determining the final disability ratings. The estimated weekly benefits coefficients are for both the impairment and disability equations. This is consistent with the income redistribution hypothesis; when everything else is the same, including injury severity and impairment rating, workers with lower temporary total benefits receive higher final disability ratings. Preinjury wages are not different from zero in the impairment rating model but positive and significant in the final disability rating

15 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 95 model. This suggests that, for a 10 percent increase in wages, there will be a corresponding 2.3 percent increase in impairment ratings and a 1.3 percent increase in disability ratings after controlling for injury severity. 9 In addition, a final disability rating that is based on loss of earnings rather than physical impairment results in a significantly higher (78 percent) final rating than one based solely on physical impairment. The coefficient on the sample selection correction variable (lambda or Mills ratio) is statistically significant, confirming the need to consider the probability of a claimant receiving an impairment and a disability rating. Unlike the previous research, the underlying probit model looks at the probability that a high cost claim receives both types of rating and not the probability that a lost time claim will receive a PPD award. 10 (The Probit results are presented in the Appendix.) 11 State differences are particularly revealing. The set of state dummy variables is jointly significant, confirming that state differences significantly influence the determination of final disability ratings. Table 4 presents the estimated coefficients for the state control variables for the jointly determined model with sample correction. 12 As might be expected, state differences are not as important for the physician's impairment rating once other factors are considered; only three individual state coefficients are statistically significant. State coefficients vary widely for the final disability rating model (Column 3, Table 4). Seven of the individual state parameters are statistically significant with the largest coefficient observed for Tennessee (1.019). The implication is that, even after controlling for injury severity, other demographic characteristics, and the impairment rating, there is tremendous interstate variation in final disability ratings. 9 Wages are generally a measure of opportunity cost and are typically negative in duration models. Here, we are looking at the impact on ratings and disability award; thus, the opportunity cost argument is not relevant. 10 The Mills ratio is entered positively into the regressions; the coefficient may be interpreted directly. 11 The Probit models are estimated using dummy variables that represent the insurers, which serve to properly, identify the system of equations. 12 The state of New Hampshire was used as the base.

16 96 The Journal of Risk and Insurance Table 4 State Variable Coefficients and t-statistics (New Hampshire is the base.) (N=4265) Physician's Impairment Rating Final Disability Rating Coefficient t-ratio Coefficient t-ratio Arkansas ** * Georgia Kentucky * Missouri 0.270* * North Carolina * South Carolina * Tennessee * Virginia Wisconsin * * * Significant at.01 ** Significant at.05 *** Significant at.10 For example, given the "average" claim profile 13 of a 38-year old man with a back strain injury treated with surgery, who retains an attorney, disputes his indemnity benefits, receives a physician's impairment rating of 12.4 percent of the whole body, and has a final disability rating that is based on physical impairment (i.e., not based on loss of earnings), the estimated parameters from the regression model predict that the claim will receive a final disability rating that varies from 9.5 percent of the whole body in New Hampshire to 26.4 percent in Tennessee (see Table 5). Table 5 Estimated Final Disability Ratings By State For Average Claim Profile Predicted Final Disability Rating by State Arkansas 18.4 Georgia 9.2 Kentucky 16.5 Missouri 16.4 New Hampshire 9.5 North Carolina 16.5 South Carolina 19.0 Tennessee 26.4 Virginia 9.8 Wisconsin The mean values used in the prediction are in some cases the means of variables in logarithmic form and may not correspond exactly with the sample means in Table 3.

17 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 97 Model 2: Effects of Final Disability Ratings on Total Indemnity and Permanent Partial Benefit Costs Models are then estimated to measure the influence of the final disability rating on the dollar amount of the permanent partial benefit award and the total indemnity benefit award. As expected, these models indicate that the final disability rating has a significant effect on both permanent partial and total indemnity benefit costs. Since the final disability rating and claim costs are in some fashion jointly determined, simultaneous equation models are estimated. The effect of the final disability rating on the total permanent partial benefit cost and on the total indemnity benefit cost are modeled as two-stage least squares with simultaneous estimation of the impairment rating, the final disability rating, and the total permanent partial benefit cost. We also simultaneously estimate the impairment rating, the final disability rating and the total indemnity benefit cost. The model estimates are presented in Table 6. Once again, the coefficient on the sample selection correction variable for the permanent cost model is insignificant, indicating that it is necessary to consider the probability that a claimant receives both an impairment rating and a disability rating. The first cost model (Column 2, Table 6) estimates the effects of selected variables on the permanent partial benefit award (corrected for sample selection bias). The second model (Column 3, Table 6) estimates the effects on the total indemnity benefit award. The regressions include the final disability rating; the claimant's age and preinjury wage, claimant attorney representation, dispute variables, and whether the claimant returned to work prior to claim closure. The models also include control variables for state and accident year (the fixed effects). The final disability rating along with the claimant's preinjury wage are expected to have the greatest effect on the permanent partial and total indemnity cost. This is because of the typical method employed to calculate a permanent partial benefit award (a percentage of the preinjury wage times the disability rating times the number of weeks in the schedule). Attorney involvement and the presence of disputes are also expected to be positively correlated as they may affect additional amounts added to the final award, such as "incentives to settle." State effects are also important, since state statutes vary regarding wage replacement rates and the number of weeks awarded for each type of impairment. The results from the models show that, as expected, the final disability rating has a significant effect on both permanent partial benefit costs and total indemnity costs. Therefore, the factors shown to have an impact in the determination of final disability ratings also will have an effect on permanent partial benefit and total indemnity benefit costs, resulting in significant disparities in benefit costs among claimants with similar kinds of injuries and preinjury wages. For example, female workers receive lower disability ratings, on average, and therefore lower benefits than their male counterparts, when comparing similar injuries as measured by the treating physician's impairment rating. Older workers appear to benefit; age is positively correlated with final disability award, again when comparing similar injuries as measured by a physician's impairment rating.

18 98 The Journal of Risk and Insurance Also, as hypothesized, claimant attorney involvement and claim disputes are positively correlated with disability ratings and claim costs, suggesting that claimants are able to increase the size of their permanent partial benefit award by investing in legal council and disputing the indemnity award.

19 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 99 Table 6 Results of Two Stage Least Squares Regression (with Sample Selection Correction) for Final Disability Rating and Total Permanent Partial Cost, Total Indemnity Cost. N = 3982 Variable Total Permanent Partial Cost Total Indemnity Cost Coefficient t. Statistics Coefficient t. Statistics Final Disability Rating 0.551* * Physician's Impairment Rating Wage 0.024* * Weekly Benefit (Temporary Total) 0.454* * Age 0.051*** Male Attorney Representation 0.098* * Dispute Over Indemnity Benefits Dispute Over Medical/Rehab. Benefits 0.098* * Loss of Earnings Base Lump Sum Award Total Medical Cost Surgery Pre-existing Condition Release to Work Return to Work 0.113* * Sprain or Strain Part of Body Controls No No No No State Controls Yes Yes Yes Yes Accident Year Controls Yes Yes Yes Yes Lambda 0.209* F statistic 78.9* * Adjust R * Significant at.01 ** Significant at.05 *** Significant at.10

20 100 The Journal of Risk and Insurance The results from the cost models suggest that a 10 percent increase in final disability rating increase permanent partial benefit costs by about 5.5 percent and total indemnity costs by approximately 3.9 percent. Other findings are consistent with the prior results and hypotheses. Wages are positively correlated with costs but negatively correlated with the disability rating. Attorney involvement increases both the disability rating (10.3 percent) and the total permanent partial award (11.9 percent). Formal indemnity disputes also increase disability awards and permanent partial costs (by 7 percent and 10 percent, respectively). 14 However, although all of the models are robust with the key variables statistically significant, it is again notable that the adjusted R 2 s are not higher, given legislative intent and the statutory structure of indemnity benefits. As mentioned above, the final disability-rating model explains only approximately 59 percent of the variation in the data, while the permanent partial cost model explains only 33 percent. This is further collaboration of how "noisy" the process is. The simultaneous cost equations are also estimated separately for each state in the sample in order to observe the variation in the estimated impact of both the impairment and the disability ratings on costs. The results show a significant difference by state, as shown by the estimated coefficients for the final rating for the separate state cost models shown (see Table 7). For example, the estimated coefficients vary from in South Carolina to in Kentucky for the permanent partial cost model, suggesting that permanent partial costs in South Carolina are affected more by disability ratings than those in Kentucky. Stated differently, not only does the determination of disability ratings vary by state after controlling for injury severity, the physician's impairment rating, and demographic characteristics, but also an increase in the disability rating itself has a differential impact on the final claim costs. These findings directly illustrate the difficulties in designing equitable permanent disability systems. Even taking into consideration that states may use PPD benefit design to redistribute income, moral hazard incentives, rent seeking activities, and other random circumstances have a significant impact on the determination of PPD benefits. CONCLUSIONS Compensating permanent partial injuries remains one of the most complicated aspects of workers' compensation insurance. Despite widespread concern over the adequacy and equity of permanent partial benefits, little research examines the factors that influence the cost of permanent partial workers' compensation claims. Using a new and rich data source on workers' compensation claims, this study 14 For continuous variables that enter the model in the natural logarithmic form, the coefficients represent the percentage change in the outcome measure given a percentage change in the individual explanatory variable. However, for the qualitative or dummy variables that enter the model as either 0 (no) or 1 (yes), the elasticity is measured as e b - 1.

21 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 101 explores these factors. The results show that impairment ratings are only one of a variety of factors that systematically influence the size of a final disability award. Specifically, even for cases with benefits awarded for noneconomic loss, in addition to the treating physician's determination of physical impairment, the determination of the degree of permanent disability appears to take into account factors such as age, sex, preinjury wage, weekly temporary total benefits, and whether an attorney is involved in the case. Moreover, even after these other factors are considered, a less than one-to-one relationship exists between impairment and final disability ratings, which might be expected. Table 7 Estimated Impact of the Final Disability Rating on Costs by State (Percent Increase in Benefit Costs Given One Percent Increase in Final Disability Rating) Permanent Partial Costs Total Indemnity Costs Percent Absolute Percent Absolute Increase t Ratio Increase t Ratio Arkansas 0.405* * Georgia 0.662* * Kentucky 0.102* Missouri 0.153* ** New Hampshire 0.557* * North Carolina 0.488* * South Carolina 0.729* * Tennessee 0.580* * Virginia 0.309* * Wisconsin 0.433* * *Significant at.01; **Significant at.05; ***Significant at.10 Although previous research suggests that impairment ratings do not adequately compensate injured workers for non-economic loss, it is clear that state systems have developed procedures that allow for additional compensation. However, equity concerns still remain. The empirical findings suggest that the same injuries as measured by a physician impairment rating may receive very different final disability ratings and thus have very different claim costs as a result of factors unrelated to the injury. Interstate differences also are revealing: Predicted final disability ratings vary significantly across states, even after controlling for injury severity, other demographic characteristics, and the

22 102 The Journal of Risk and Insurance impairment rating. As observed during the reports of the National Commission on State Workmen's Compensation Laws in the early 1970s, the equitable distribution of permanent disability benefits is one of the most complex and controversial issues affecting workers' compensation. The high levels of disputes and litigation are continuing signals of the problems. Even after many years of often-significant reforms, these issues remain today. APPENDIX Table 8 Probit Model Probability That Claim Has Impairment Rating and Disability Rating (for Sample Selection Correction) Variables Estimated T-Ratio Coefficients Insurer # * Insurer # ** Insurer # Insurer # * Insurer # ** Insurer # * Insurer # ** Insurer # * Insurer # *** Total Medical Cost 0.126* Total Indemnity Cost * Loss of Earnings Base State Controls Yes * Significant at.01 ** Significant at.05 *** Significant at.10 REFERENCES Berkowitz, M. and J. Burton, 1987, Permanent Disability Benefits in Workers' Compensation Kalamazoo, Mich.:. W.E. Upjohn Institute for Employment Research. Boden, L., 1992, Dispute Resolution in Workers' Compensation, Review of Economics and Statistics, 74: Boden, L. and R. Victor, 1994, Models for Reducing Workers' Compensation Litigation, Journal of Risk and Insurance, 61(3): Borba, P., and D. Appel, 1987, The Propensity of Permanently Disabled Workers to Hire Lawyers, Industrial and Labor Relations Review, 40(3):

23 Factors Affecting Permanent Partial Disability Ratings In Workers Compensations n 103 Durbin, D., N. Helvacian, and R. Butler, 1995, Attorney Involvement in High Cost Workers' Compensation Cases, NCCI Working Paper, Boca Raton, Florida. Durbin, D., and R. Butler, 1998, Prevention of Disability from Work Related Sources, in New Approaches to Disability in the Workplace, Terry Thomason, John Burton Jr. and Douglas Hyatt, eds., (Madison, WI: IRRA Press). Greene, W., 1981, Sample Selection Bias as a Specification Error: Comment, Econometrica, 49(3): Halvorsen, R. and R. Palmquist, 1984, The Interpretation of Dummy Variables in Semilogarithmic Equations, American Economic Review, 70(3): Heckman, J., 1979, Sample Selection Bias as a Specification Error, Econometrica, 47(1), Krueger, A. B. and J. F. Burton, 1990, The Employer's Costs of Workers' Compensation Insurance: Magnitudes, Determinants and Public Policy, Review of Economics and Statistics, 81: National Council on Compensation Insurance, , NCCI Annual Statistical Bulletin (Boca Raton, Fla.: NCCI). Sinclair, S. and J. Burton, 1994, Measuring Non-Economic Loss: Quality of Life Values versus Impairment Ratings, Workers' Compensation Monitor. Thomason, T., 1993a, Permanent Partial Disability in Workers' Compensation: Probability and Costs, The Journal of Risk and Insurance, 60: Thomason, T., 1993b, The Transition from Temporary to Permanent Disability: Evidence from New York State, in, Durbin and Borba, Eds. Workers' Compensation Insurance: Claim Costs, Prices, and Regulation Boston: (Kluwer Academic Publishers). Worrall, J., D. Durbin, D. Appel, and R. Butler, 1993, The Transition from Temporary Total to Permanent Partial Disability: A Longitudinal Analysis, in: Durbin and Borba, eds., Workers' Compensation Insurance: Claim Costs, Prices, and Regulation, Boston: (Kluwer Academic Publishers).